newsroom.sgit.ai / world news day / piece 8
World News Day 2026 · beta

The corpusThe licenceThe 21The aggregateThe graphSourcesHow to reach themThe vaultMethod

Piece 8 of 21 · World News Day 2026

Truth in an AI World

33 sentences, typed by a published formula and anchored to bytes we hold. The piece itself is not here — read it at worldnewsday.org, where its authors put it.

Published 2026-09-27 · 649 words · frozen copy as served and as the publisher's API returns it · SHA-256 94bded9b048e5f09fb7b069f…

Beta, agent-produced. The fetching, extraction, classification and drafting here are done by software agents against frozen bytes. Nothing on this page is a legal opinion, and there has been no legal review. If you are deciding whether you may republish one of these pieces, read the terms on the piece itself and ask the publisher — that is the point we are making.

We hold all twenty-one pieces and republish none of them. Every article is linked to its original on worldnewsday.org. What is published here is the description: who wrote it, under what terms, what it links to, what words it contains. The prose belongs to the people who wrote it, and the gate on this section fails the build if twelve consecutive words of any piece appear on any page we generate.

Every number walks back to bytes we hold. Each page was fetched once, frozen to a dated snapshot in this repository and hashed with SHA-256. The counts are re-derived from those bytes by a second program before anything ships. The register → · The method →

Who wrote it

Michael Miller

Executive Chairman, News Corp Australasia

Route: News Corp Australasia contact page

What the piece does

Each sentence is classified by the first pattern that matches it, in published order. A type says this sentence has this shape — not that it is true, not that the author believes it, and an evaluation is not an accusation: an op-ed is supposed to evaluate. The corpus column is what makes the piece column mean anything.

ShapeHereShareCorpusWhat the type means
Question00%3%The sentence asks something. Rhetorical or real — this does not distinguish them.
Attributed statement00%2%The sentence hands the claim to somebody else. Placed SECOND, ahead of proposal and hypothesis, deliberately: it is the only class that points outside the piece, and this section's central finding is about evidence. Ordering it late would have undercounted attributions and flattered our own argument.
Proposal00%1%The sentence says something ought to be done. The heart of an op-ed, and the thing this corpus has least of relative to diagnosis.
Hypothesis39%7%The sentence supposes rather than asserts: a condition, a possibility, a consequence that has not happened yet.
Quantified observation13%5%The sentence carries a number, a date or a count. The kind of statement a reader could in principle check — if the piece said where it came from.
Evaluation13%4%The sentence judges. An op-ed is supposed to; this is not a criticism, it is a measurement of form.
Assertion2885%77%Everything else: a declarative sentence stated flat, in the author's own voice, with no hedge, no number, no source and no explicit judgement. The default, and in this corpus the largest class by a wide margin.

Where it touches the rest

The vocabulary this piece shares with pieces that never cite it. This is the connective tissue of the corpus: twenty-one arguments that arrive at the same words independently.

Themes, from the separate theme lexicon: Artificial intelligence, Audience, attention and avoidance, Democracy and civic life, Investigative reporting, Local and community news, Misinformation and propaganda, Money and the business model, Platforms and distribution, Trust and credibility, Truth and facts.

What this connects to in our own argument

By a published rule over the words the piece uses — a row appears because the piece uses those terms and we have written about them. It does not claim the author would agree with us, and several of them would not.

Provenance is the product

A claim is only as good as the chain a reader can walk back from it. The corpus asks the public to trust journalism; this is the mechanism by which they could check it instead.

Because this piece uses: facts, trust, verification. Read it →

Corrections must propagate

A correction that never reaches what it disproved is not a correction. Nobody in the corpus proposes a mechanism for this, and it is the one that would matter most.

Because this piece uses: disinformation, misinformation, truth. Read it →

The thesis: sell the graph

A story is a graph and an article is one projection of it. If a model is going to read your work anyway, the question is what it reads and on what terms.

Because this piece uses: artificial intelligence, platform. Read it →

What it offers the reader

Every sentence

An anchor is at most 8 words, verbatim, so you can find the sentence in the original. The sentence itself is not stored here, and the build fails if twelve consecutive words of this piece appear on any page we generate.

#ShapeAnchorTerms
1assertion“By Michael Miller .…”
2assertion“Today marks World News Day, when newsrooms worldwide…”trust
3assertion“This year, that point needs making more urgently…”
4assertion“Democracies run on a shared set of facts.…”facts
5assertion“Courts, elections, markets, even ordinary arguments at the…”
6assertion“Strip that away, and you don’t get healthy…”
7assertion“You get a society where people shout past…”
8assertion“That shared footing is now under real pressure,…”
9evaluation“Artificial intelligence can be an extraordinary force for…”artificial intelligence
10hypothesis“But the same technology is equally capable of…”misinformation
11assertion“Even the leaders of the companies building these…”
12assertion“In Australia in the last week OpenAI admitted…”artificial intelligence
13assertion“This followed an earlier admission its GPT-5.6 Sol…”
14hypothesis“If the makers of the technology are worried,…”
15assertion“Part of what makes this moment different is…”disinformation
16assertion“A recent academic study out of Europe put…”artificial intelligence
17assertion“The tools sound authoritative regardless of whether what…”
18assertion“These AI models are engineered to maximise engagement,…”artificial intelligence, facts
19assertion“That gap between feeling informed and being informed,…”
20assertion“The human toll on the dilution of truth…”truth
21assertion“Scammers are using AI-generated video and voices to…”artificial intelligence
22assertion“Criminals are manufacturing fake identities and fraudulent documents…”
23assertion“In Australia alone, scams and fraud enabled by…”community
24assertion“There is a quieter cost too, one that’s…”
25assertion“With more revenue flowing to tech platforms, local…”platform
26quantified“A 2024 study measured what follows: when local…”community, local news
27assertion“That’s not a coincidence – it’s what happens…”
28assertion“None of this is unfixable, but it won’t…”
29hypothesis“A well-functioning society still needs people whose job…”verification
30assertion“So today, on World News Day, the ask…”journalism
31assertion“Trust it over a slick, unsourced summary.…”trust
32assertion“Because as the tools for manufacturing fiction keep…”facts
33assertion“Michael Miller : Executive Chairman, News Corp Australasia…”

For an agent

This page is a projection of ../data/claims.json (filter by article == "truth-in-an-ai-world"), ../data/corpus.json and ../data/terms.json. Each claim carries a type from a published formula, an offset into the frozen prose and an anchor of at most 8 words. The article text is in none of them — follow url to the publisher. A claim type is a statement about the SHAPE of a sentence and about nothing else.